ARIMA model

Inspiration

I am a 3rd Mathematics Student at Imperial College London. I am currently attending the Time Series course from my degree. I am motivated to see how the models that I am learning from the lectures can be applied to real data. This inspired me to carry out an analysis on Quandl's data which can be then used to determine whether it is suitable for forecasting future prices.

Data used

I extracted the data from Quandl. I used Bank of England's Official Statistics on the Forward Exchange Prices of USD into Sterling. For analysing the time series with the ARIMA model, I used the linear returns on the 1-month Forward Exchange Prices of USD into Sterling.

Challenges I ran into

All my teammates snaked on me. Bless them.

What I learned

At first, I decided to apply the ARMA(p,q) model that I learnt from lectures to my data. However, it suggests that my time series is not stationary. This led me to research and discover the ARIMA(p,d,q) model that enables me to transform non-stationary white noise to stationary by first-differencing (d=1). Hence, analysis on the time series can be carried out.

Built With

• python
• quandl